Improving the measurement resolution in BOTDR sensor with optimized wavelet denoising strategy

Area of Science:

• Optical Engineering
• Signal Processing
• Fiber Optic Sensing

Background:

• Signal averaging in Brillouin optical time domain reflectometry (BOTDR) creates a trade-off between measurement time and resolution.
• Existing methods struggle to achieve high resolution without extensive averaging, limiting real-time applications.

Purpose of the Study:

• To develop an optimized post-processing strategy integrating wavelet denoising (WD) with Lorentzian curve fitting (LCF) to mitigate the signal averaging trade-off in sub-meter differential cross-spectrum BOTDR (DCS-BOTDR).
• To evaluate the performance of WD functions (Symlet, Daubechies, Coiflet, Biorthogonal Spline) with a 4-level decomposition on a six-core CPU using the single-program-multiple-data (SPMD) paradigm.

Main Methods:

• Integration of wavelet denoising (WD) functions with traditional Lorentzian curve fitting (LCF).
• Implementation of a 4-level decomposition using various WD functions (Symlet, Daubechies, Coiflet, Biorthogonal Spline).
• Parallel processing on a six-core CPU utilizing a single-program-multiple-data (SPMD) paradigm for computational efficiency.

Main Results:

• Achieved a 2.7 MHz Brillouin frequency shift (BFS) resolution with only 21,000 averages, compared to 56,000 averages for LCF alone.
• Reduced required signal averages by 2.7 times.
• Preserved 0.4 m spatial resolution and improved temperature resolution to 3°C over 1.21 km fiber with 14,000 averages.
• Accelerated data processing speed by up to 4.8 times using a parallel multicore architecture.

Conclusions:

• The proposed LCF + WD method offers a significant improvement in resolution and measurement time for DCS-BOTDR.
• The training-free, straightforward strategy provides comparable BFS and temperature resolutions to machine learning methods without extensive datasets.
• The parallel implementation is highly beneficial for real-time distributed sensing applications with constrained computational resources.